timeseries.whiten: don't produce NaN when zeros in asd
Created by: andrew-lundgren
timeseries.whiten: don't produce NaN when zeros in asd
I ran into trouble in q_transform due to NaNs in the whitened data. This was happening because some values of the asd were zero, making the inverse blow up. The patch deals with this the simplest way, just replacing all zero values with the minimum positive value.